A-Novel-Recursive-Network-for-Irony-Detection-in-Tweets

URL: https://github.com/guangyizhangbci/A-Novel-Recursive-Network-for-Irony-Detection-in-Tweets

Description: Course Project (ELEC 880 @ Queen’s University)

Project Overview

This project implements a novel recursive network for irony detection in tweets. The dataset used for training and evaluation is from the SemEval-2018 Task 3, which focuses on detecting irony in English tweets. The project is based on PyTorch and includes various models and training strategies.

Dataset:

The dataset used in this project is part of the SemEval-2018 Task 3. It contains English tweets labeled for irony detection.

Dataset Usage:

The repository provides instructions on how to use the dataset with different models, including recursive networks and BERT-based models. The models can be trained by running the corresponding training scripts.

Dataset Statistics:

Details regarding the dataset size or number of samples are not explicitly mentioned in the provided information.

Training Methods:

The project involves several approaches:

1. Recursive Network Training: Run the models/task3/train for the NTUA (National Technical University of Athens) implementation.
2. BERT-based Model: Run the /Vandad/try-bert script to experiment with BERT for irony detection.
3. Embeddings: Download NTUA embeddings from their GitHub repository and place them under the “embeddings” folder for model training.

Results:

The results of the experiments can be found in the /out/experiments/ folder. Specific evaluation metrics (accuracy, precision, recall, etc.) are not detailed in the provided information.